Characteristic and intermingled neocortical circuits encode different visual object discriminations

Abstract

Synaptic plasticity and neural network theories hypothesize that the essential information for advanced cognitive tasks is encoded in specific circuits and neurons within distributed neocortical networks. However, these circuits are incompletely characterized, and we do not know if a specific discrimination is encoded in characteristic circuits among multiple animals. Here, we determined the spatial distribution of active neurons for a circuit that encodes some of the essential information for a cognitive task. We genetically activated protein kinase C pathways in several hundred spatially-grouped glutamatergic and GABAergic neurons in rat postrhinal cortex, a multimodal associative area that is part of a distributed circuit that encodes visual object discriminations. We previously established that this intervention enhances accuracy for specific discriminations. Moreover, the genetically-modified, local circuit in POR cortex encodes some of the essential information, and this local circuit is preferentially activated during performance, as shown by activity-dependent gene imaging. Here, we mapped the positions of the active neurons, which revealed that two image sets are encoded in characteristic and different circuits. While characteristic circuits are known to process sensory information, in sensory areas, this is the first demonstration that characteristic circuits encode specific discriminations, in a multimodal associative area. Further, the circuits encoding the two image sets are intermingled, and likely overlapping, enabling efficient encoding. Consistent with reconsolidation theories, intermingled and overlapping encoding could facilitate formation of associations between related discriminations, including visually similar discriminations or discriminations learned at the same time or place.

Keywords: Activity-dependent gene imaging; Characteristic circuits; Neocortical circuits; Overlapping encoding; Visual discrimination learning.

Fig. 1

In rats that received PkcΔ, during performance of an image set learned after gene transfer, the genetically-modified circuits contain bilaminar patterns of active neurons. (A) A time line showing the experimental design. Rats were trained on a control image set, gene transfer was performed, rats were retested on the control image set, trained on one new image set for 10 sessions, and sacrificed. Activity-dependent gene imaging (c-fos-IR) was performed. (B–F) Top row, [] vs. +. (G–K) bottom row,/ vs. \. (B, C, G, H) Near an injection site, low power (B, G), or high power (C, H); arrows, c-fos-IR cells. (D, I) POR cortex, away from an injection site. (E, J) Serial section reconstructions showing the positions of the active neurons in the genetically-modified circuits. (F, K) Distances from the surface of POR cortex to the active neurons. Scale bars: (B, D, G, I) 100 µm; (C, H) 25 µm; (E, J) 1 mm.

Fig. 2

The distributions of distances from the surface of POR cortex to active neurons for individual hemispheres in rats that received PkcΔ followed by learning [] vs.+. For each hemisphere, serial section reconstructions showing the positions of the active neurons in the genetically-modified circuits are on the left, and the distances from the surface of POR cortex to the active neurons are on the right. One rat is shown in each row.

Fig. 3

The distributions of distances from the surface of POR cortex to active neurons for individual hemispheres in rats that received PkcΔ followed by learning / vs. \. For each hemisphere, serial section reconstructions showing the positions of the active neurons in the genetically-modified circuits are on the left, and the distances from the surface of POR cortex to the active neurons are on the right. One rat is shown in each row. For one rat (bottom row), there were technical difficulties with the histology for the right hemisphere.

Fig. 4

Specific image set are encoded in bilaminar patterns of active neurons. For each image set and condition, the distribution of the distances from the surface of POR cortex to the active neurons (numbers of hemispheres and rats per group are in the text).

Fig. 5

In wt rats, during learning, preferentially-activated circuits in POR cortex contain bilaminar patterns of active neurons. (A–D) Top row, [] vs.+. (E–H) bottom row, / vs. \. (A, E) Sections containing preferentially-activated circuits, low power; arrows, c-fos-IR cells. (B, F) POR cortex, sections distant from the preferentially-activated circuits. (C, G) Serial section reconstructions showing the positions of the active neurons in the preferentially-activated circuits. (D, H) Distributions of the distances from the surface of POR cortex to the active neurons. Scale bars: (A, B, E, F) 100 µm; (C, G) 1 mm.

Fig. 6

The distributions of distances from the surface of POR cortex to active neurons for individual hemispheres in wt rats that learned [] vs. +. For each hemisphere, serial section reconstructions showing the positions of the active neurons in the preferentially-activated circuits are on the left, and the distances from the surface of POR cortex to the active neurons are on the right. One rat is shown in each row. For one rat (bottom row), there were technical difficulties with the histology for the right hemisphere.

Fig. 7

The distributions of distances from the surface of POR cortex to active neurons for individual hemispheres in wt rats that learned / vs. \. For each hemisphere, serial section reconstructions showing the positions of the active neurons in the preferentially-activated circuits are on the left, and the distances from the surface of POR cortex to the active neurons are on the right. One rat is shown in each row. For one rat (bottom row), there were technical difficulties with the histology for the left hemisphere.

Fig. 8

In rats that received the control vector, PkcΔGG, during performance of an image set learned after gene transfer, the genetically-modified circuits contain few active neurons that are not arranged in bilaminar patterns. (A–C) Top row, [] vs. +. (D–F) bottom row, / vs. \. (A, D) Near an injection site, low power; arrows, c-fos-IR cells. (B, E) Serial section reconstructions showing the positions of the few active neurons in the genetically-modified circuits. (C, F) Distances from the surface of POR cortex to the active neurons. Scale bars: (A, D) 100 µm; (B, E) 1 mm.

Fig. 9

The distributions of distances from the surface of POR cortex to active neurons for individual hemispheres in rats that received PkcΔGG followed by learning an image set. For each hemisphere, serial section reconstructions showing the positions of the active neurons in the genetically-modified circuits are on the left, and the distances from the surface of POR cortex to the active neurons are on the right. One rat is shown in each row. (A) [] vs. +; (B) / vs. \.